But the order of attributes in the presentation of the result relation has nothing to do with physical layout of base relvars -- primarily because any relationship between the two is purely coincidental. The vast majority of useful queries will not reuse the order of attributes in base relvars, so optimizing for the unusual trivial case by prohibiting better rearrangements that could be useful for a much larger number of use cases seems rather pointless.

And what PostgreSQL does is of course an implementation detail of PostgreSQL.